Currently, distributed Raman optical fiber amplifiers (RFAs) and erbium-doped optical fiber amplifiers (EDFAs) have been widely used in communication systems. To meet a variety of requirements for application indices and functionalities, a number of control processes are used in product designs of the optical fiber amplifiers. For example, feedforward and feedback are combined to improve transient indices, and a prediction-based method is used to realize a pre-response to fast timing. In the following scenarios, an appropriate controller design is critical to achieve control functionalities and performances that satisfy the requirements:
In a situation where the control target has a plurality of parameters, it is usually difficult to achieve control performance that meets the requirements. In an EDFA amplifier, for example, gain is limited in a power mode, or power is limited in a gain mode. In such cases, a control process using a single loop can hardly overcome impact of disturbance to obtain desirable dynamic characteristics.
To optimize the system indices or to achieve certain functional characteristics in a multi-pumping optical fiber amplifier, various pumping lasers may be required to have its power controlled according to some parameters or control goals. For example, in a gain tunable distributed Raman optical fiber amplifier, power of pumping lasers of different wavelengths is determined according to a predetermined proportion to achieve control on tilt and gain; or in an EDFA amplifier, in order to obtain better optical indices, pumping power is assigned according to a fixed proportion when the input optical power and gain are within a particular range and assigned according to another proportion when the input optical power and gain are in other ranges. For such system requirements, the controller cannot be implemented directly by using a single-loop controller. A compromise is to calibrate the pumping power-current curve in manufacturing of the products, and the single loop controller is used to carry out real-time control with reference to a lookup table that refers power to current. However, such a compromise adds the cost of time and labor to carry out the calibration, and as time flies, the pumping power-current curve may change due to aging of the pumping lasers, which cause the control goals unachievable, and the amplifier cannot converge to a correct operating point.